Mir-106B-Responsive Gene Landscape Identifies Regulation of Kruppel- Like Factor Family

bioRxiv preprint doi: https://doi.org/10.1101/088229; this version posted November 17, 2016. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

miR-106b-responsive gene landscape identifies regulation of Kruppel- like factor family

Cody J. Wehrkamp Sathish Kumar Natarajan Ashley M. Mohr Mary Anne Phillippi Justin L. Mott

Department of Biochemistry and Molecular Biology, Fred & Pamela Buffett Cancer Center, University of Nebraska Medical Center, Omaha

Running title: miR-106b-responsive genes in cholangiocarcinoma

Key words: Apoptosis, biliary tract cancer, KLF2, KLF4, KLF6, KLF10, KLF11, KLF13, LKLF, lung Kruppel-like factor, miR-106a, miRNA, next-generation sequencing

Address for Correspondence: Justin L. Mott, MD, PhD Associate Professor Department of Biochemistry and Molecular Biology 985870 Nebraska Medical Center Omaha, NE 68198-5870 Tel: 402-559-3177 Fax: 402-559-6650 e-mail: [email protected]

List of abbreviations: 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide, MTT; 4′,6-diamidine-2′- phenylindole dihydrochloride, DAPI; Crosslinking, ligation, and sequencing of hybrids, CLASH; Death receptor-5, DR5; Dulbecco’s Modified Eagle’s Medium, DMEM; False discovery rate, FDR; Fetal bovine serum, FBS; Kruppel-like factor, KLF; Locked-nucleic acid, LNA; Quantitative reverse-transcription PCR, qRT-PCR; Sodium dodecylsulfate-polyacrylamide gel electrophoresis, SDS-PAGE; Untranslated region, UTR;

miR-106b-responsive gene landscape identifies regulation of Kruppel-like factor family Abstract: MicroRNA dysregulation is a common feature of cancer and due to the promiscuity of microRNA binding this can result in a wide array of genes whose expression is altered. miR- 106b is an oncomiR overexpressed in cholangiocarcinoma and its upregulation in this and other cancers often leads to repression of anti-tumorigenic targets. The goal of this study was to identify the miR-106b-regulated gene landscape in cholangiocarcinoma cells using a genome- wide, unbiased mRNA analysis. Through RNA-Seq we found 112 mRNAs significantly repressed by miR-106b. The majority of these genes contain the specific miR-106b seed- binding site. We have validated 11 genes from this set at the mRNA level and demonstrated regulation by miR-106b of five proteins. Combined analysis of our miR-106b-regulated mRNA data set plus published reports indicate that miR-106b binding is anchored by G:C pairing in and near the seed. Novel targets Kruppel-like factor 2 (KLF2) and KLF6 were verified both at the mRNA and at the protein level. Further investigation showed regulation of four other KLF family members by miR-106b. We have discovered coordinated repression of several members of the KLF family by miR-106b that may play a role in cholangiocarcinoma tumor biology.

Introduction microRNA which is vital for complementary binding to its mRNA target. The miR-106b has been established as an complementary seed-binding site is more or oncogenic microRNA with increased less conserved among targets, allowing for expression in many cancers including broad transcript-expression dampening cholangiocarcinoma [1-5], prostate [6], effects from a single microRNA. Seed- gastric [7], and hepatocellular carcinoma [8]. binding sites are highly enriched in the 3’ Some functions of miR-106b include untranslated region (UTR) of transcripts increased proliferation by miR-106b- [17]. There is also evidence that supports mediated reduction of the transcription non-canonical, non-seed mRNA target factor E2F1 [9] and the tumor suppressor interactions between microRNAs and their RB1 [10]. Additionally, miR-106b increased targets [18, 19]. The traditional seed for migration through targeting of the miR-106b is located at nucleotides 1-8 on phosphatase PTEN [11], and reduced its 5’ end, corresponding to the microRNA apoptosis by regulating the BH3-containing sequence 5’UAAAGUGC. protein Bim [12]. In this study, we sought to define the MicroRNAs function to dampen the genome-wide target set of miR-106b and expression of their targets [13-15]. This is found that miR-106b binding relied on accomplished both through reduction in sequences between bases 2-10, tolerating a mRNA transcript level as well as by G:U wobble in the seed region. We found repression of translation. While both 112 transcripts that were regulated at the mechanisms contribute to reduced target mRNA level in cholangiocarcinoma cells. protein expression, the effect of mRNA Finally, we demonstrated that members of reduction may dominate [16], in part the KLF family of transcription factors are because each message is used as a coordinately repressed by miR-106b. transcript for synthesis of many protein Comparison of our data in molecules; still, the magnitude of change in cholangiocarcinoma cells with data in Flp-In mRNA expression is low. A focused T-REx 293 human embryonic kidney- description of the binding characteristics derived cells [19] revealed that the and targets of individual microRNAs is microRNA-responsive gene set is largely feasible and desirable. cell-type specific. The seed region of a microRNA is a 7-8 nucleotide sequence at the 5’ end of the

Materials and Methods Sequences were analyzed by the Bioinformatics and Systems Biology Core at Cell lines: Human cholangiocarcinoma cell the University of Nebraska Medical Center lines were previously derived from a female (UNMC) using the Tuxedo protocol [23]. patient with metastatic gallbladder cancer, Reads were mapped to the human genome Mz-ChA-1 [20], or a male patient with using Top Hat. Transcripts were assigned combined histologic features of by Cuff links and the transcriptome defined hepatocellular carcinoma and using Cuffmerge. Finally, differential cholangiocarcinoma, KMCH [21]. BDEneu expression between miR-106b and LNA- rat cholangiocarcinoma cells were derived 106b was calculated with Cuffdiff. from primary Fisher 344 rat cholangiocytes [22]. H69 cells are a non-malignant Using a false discovery rate (FDR)- immortalized cholangiocyte cell line. corrected p value, genome-wide transcripts Cholangiocarcinoma cells were grown in were ranked from most significant to least. Dulbecco’s Modified Eagle’s Medium This ranked list of genes was submitted to (DMEM) supplemented with 10% fetal the SylArray online server to detect bovine serum (FBS), insulin (0.5 µg/mL) enrichment of microRNA seed-binding sites and G418 (50 µg/mL). H69 cells were [24]. grown in DMEM supplemented with 10% miR-106b binding site analysis: RefSeq FBS, insulin (5 µg/mL), adenine (24.3 sequences for the 129 significant genes µg/mL), epinephrine (1 ug/mL), tri- (112 decreased and 17 increased) were iodothyronine (T3)- transferrin (T), [T3- 2.23 collected in Fasta format and analyzed by ng/ml, T-8.19 µg/mL], epidermal growth direct search for the known miR-106b factor (9.9 ng/mL) and hydrocortisone (5.34 binding site as well as subjected to k-mer µg/mL). analysis to generate a k table of all possible Cells were transfected either with mature 7-mer and 8-mer sequences and their miR-106b mimic (Life Technologies), frequency in this gene set. Several genes locked-nucleic acid (LNA) antagonist to had more than one RefSeq sequence, so miR-106b, LNA-106b (Exiqon), or LNA the final size of this Fasta file was 191 Negative Control A (Exiqon) for 24-48 hours transcripts. For comparison, 1,000 random using Lipofectamine RNAiMAX at a final sets of 191 transcripts each from the oligonucleotide concentration of 20 nM. RefSeq database were generated. miR-106b levels were quantified by TaqMan RNAHybrid [25] was used to identify the Small RNA assay (Life Technologies). single best miR-106b predicted binding site RNA-Seq: High quality total RNA was for each decreased transcript based on extracted from Mz-ChA-1 cells transfected minimum free energy hybridization. The with either miR-106b (n=3) or LNA-106b resulting miR-106b-binding sites were (n=3) for RNA-Seq using RNeasy Mini Kit manually curated for discovery of ungapped (Qiagen). The quality of the RNA was motifs using MEME Suite [26]. confirmed by RNA Integrity Analysis using Quantitative reverse-transcription PCR: an Agilent Bioanalyzer. RNA sequencing Total RNA for qRT-PCR was isolated using libraries were prepared by the UNMC Next TRIzol Reagent (Life Technologies) and Generation Sequencing Core Facility quantified using SYBR Green DNA binding beginning with 1 µg of total RNA and a (Roche). Primer pairs are listed in Table 1. TruSeq RNA Prep V2 (Illumina Inc, San Diego, CA). Samples were sequenced MicroRNA biotinylation: Biotinylation of Cel- using the Illumina HiSeq 2000 system with miR-67 and miR-106b was performed as 100 bp paired-end reads. An average of described [27]. Mature microRNA for C. 76.99 million reads per sample were elegans miR-67 (control) and H. sapiens collected (range 58.99-87.00 million). miR-106b (both leading and passenger

strand) were from Integrated DNA Tech, 10% of cell lysate was used for input RNA. Coralville, IA. Leading strand of both Cel- Streptavidin-bead bound RNAs were miR-67 and miR-106b was biotinylated at washed five times with 20 mM Tris, pH 7.5, the 3’ end using T4 RNA ligase (pierce RNA 0.5 M NaCl and 1 mM EDTA. After pull 3’ End Biotinylation kit, Thermo Scientific , down, RNA was isolated using mirVana kit. #20160). Biotinylated RNA was quantified Relative expression of IL8, KLF2 and α by dot blot on hybond N+. An equal amount Tubulin1A was analyzed by qRT-PCR. 18S of biotinylated mature microRNA (leading was used as a control RNA. strand) was mixed with its respective Immunoblotting: Lysates were probed for passenger strand RNA. KLF2 (Aviva Systems Biology), KLF4 (Cell Pull down of biotinylated RNA: Mz-ChA-1 Signaling), KLF6 (Santa Cruz cells were transfected with 50 nM of Biotechnology), KLF10 (abcam), DR5 (Cell biotinylated microRNA using Lipofectamine Signaling) or actin (Sigma). RNAiMAX in triplicate. After 24 hours, cells Apoptosis and proliferation: Treated cells were lysed (20 mM Tris, pH 7.5, 100 mM were assayed for apoptosis by nuclear KCl, 5 mM MgCl2, 0.3% NP40, 50 U of morphology, and for cell proliferation by 3 RNase OUT and complete protease (4,5 dimethylthiazol-2-yl)-2,5- inhibitor) and incubated on ice for 10 diphenyltetrazolium bromide (MTT) assay, minutes. 90% of cell lysate was incubated as described [28]. All experiments were with streptavidin magnetic beads (New repeated at least three times. England Biolabs) for 6 hours at 4⁰C and

Results and the corresponding 8-mer binding site is 5’GCACUUUA. Of the 112 mRNAs with RNA-Seq to determine miR-106b targets decreased expression, 81 had either an 8- To determine miR-106b targets, we mer or a 7-mer binding site (72.3%), and 31 experimentally manipulated miR-106b levels had at least one perfect 8-mer binding site in human Mz-ChA-1 cholangiocarcinoma (27.7%; Figure 1C). Conversely, of the 17 cells (Figure 1A). Total RNA from each mRNAs with increased expression in cells condition, miR-106b and LNA-106b, was transfected with miR-106b, only one mRNA sequenced. Messages with decreased contained a 7-mer binding sequence (5.9%) counts in miR-106b-transfected samples and none had a perfect 8-mer site (Figure compared to LNA-106b-transfected samples 1C). The enrichment of seed-binding sites were defined as repressed by miR-106b, suggested that our technique was effective while mRNAs that had increased counts in detecting miR-106b targets. were defined to be increased by miR-106b. SylArray analysis More genes were repressed by miR-106b than increased, with 112 mRNAs repressed The entire gene set was analyzed by and 17 increased (Figure 1B). Based on SylArray [24] to detect enrichment of the RefSeq sequences, we searched the microRNA binding sites. We confirmed the two data sets-- genes with decreased enrichment of the miR-106b 7-mer m8 expression and those with increased seed-binding site, 5’GCACTTT (red line) in expression-- for the miR-106b seed-binding the sequences that were significantly site. Specifically, we sought perfect 8-mer altered by miR-106b (to the left on the plot; binding sites and either 7-mer with a match Figure 1D). Notably, this analysis also at position 8 (m8) or 7-mer with an A revealed enrichment of a 7 nucleotide opposite position 1 (A1). The seed sequence complementary to nucleotides 3-9 sequence of miR-106b is 5’UAAAGUGC of miR-106b (5’AGCACTT, blue line).

Sequence determinants of miR-106b sequences were found in the top 10 targeting percentile (Figure 2C&D). Using previously published data [19], we To determine the tolerance for single base determined that 76.2% of miR-106b target differences within the seed sequence, we interactions employed 7 or more started from a perfect 1-8 sequence and consecutive miR-106b bases in the systematically searched for one-off variants. microRNA-mRNA binding hybrid. Because a For example, the 1-8 binding site high proportion of interactions contained at 5’GCACTTTA was observed 63 times while least 7 consecutive bases, we determined the sequence 5’ACACTTTA occurred 26 which of the miR-106b 7- or 8-mer binding times. The sequence 5’GCATTTTA contains sites were favored in our target gene set. a single base difference opposite position 5 We searched the 112 significant mRNA of miR-106b and was observed 40 times sequences for all 65,536 possible 8-mer (96th percentile). Compare this result to the sequences and plotted the frequency of sequence 5’GCAGTTTA which was occurrence of each 8-mer (count) versus observed only 13 times (66th percentile). We the number of 8-mers in each bin (Figure plotted the counts for each of these one-off 2A). Approximately 50% of all possible 8- sequences as a function of the position in mers were observed with a frequency of the miR-106b binding site (Figure 2E). The between 1 and 10 occurrences, while 5,020 most tolerance was observed opposite 8-mers were not observed at all (count = 0). position 1 of miR-106b. The next most Notably, the perfect 8-mer binding site tolerated substitution was a T at position 5, 5’GCACTTTA was present 63 times and the which would result in a G:U wobble base overlapping 8-mer (2-9) was present 81 instead of a G:C pair. times. The two most over-represented 8- The raw count for each sequence may not mers were 5’AAAAAAAA (1,987 times) and indicate over-representation but rather may 5’TTTTTTTT (830 times). The analogous k- indicate frequently encountered 8-mers mer data set was developed for all possible (e.g., 5’AAAAAAAA within the poly-A tail). 7-mer sequences (16,384 possible 7-mers) Thus, we sought to determine the statistical within the significant gene set (Figure 2B). significance of a number of sequences Again the most common sequences were commonly observed in our set of 129 5’AAAAAAA and 5’TTTTTTT. Forty-nine 7- significant genes. This set contained 191 mers were absent in the data set. The sequences due to some genes having sequence complementary to bases 1-7 of multiple forms. To correct for natural miR-106b was observed 121 times, the 2-8 frequency variation, we generated 1,000 sequence was observed 170 times, and the additional sets of mRNAs each containing 3-9 sequence was observed 147 times. 191 random transcript sequences and Each of the possible miR-106b- compared the frequency of the miR-106b complementary 8-mers were tiled from 5’ to binding sequence and related sequences. 3’ (i.e., 1-8, 2-9, 3-10, etc.). The 8-mer The 8-mer sequence 5’GCACTTTA was opposite bases 2-9 was somewhat more over-represented in our gene set and this frequent that the 1-8 complement (Figure finding was highly statistically significant (p 2C). A similar plot of the frequency of all = 0). Similarly, each of the 7-mer sequences complementary 7-mers tiled from 5’ to 3’ was statistically different in our set (p = 0). showed that over-represented sequences The two C’s within the miR-106b binding again favored the 5’ end of miR-106b, sequence were reassigned as T’s to search especially the classical seed (Figure 2D). for seed-binding sites including a G:U We used an unrelated, control microRNA wobble, C5 and C7. The wobble at the 5th let-7a; the same plot of tiled 8- and 7-mer position, 5’GCATTTTA, was highly sequences showed that none of these significant, as was the double wobble, 5’GTATTTTA. Tolerance for a G:U wobble

at position 7 only, 5’GTACTTTA, was not as anchored by G:C pairing in and near the highly significant (p = 0.022, Figure 2F). seed. RNA hybrid analysis Genome-wide target identification Although the perfect complement is likely Next, the identified transcripts in our RNA- preferred, our data indicated that related Seq data set were plotted by change in sequences were also commonly expression versus statistical significance overrepresented. We sought to identify the (volcano plot). The relative change in characteristics of the most expression between miR-106b and LNA- thermodynamically stable miR-106b binding 106b samples was plotted as log2 of the site in our significantly-repressed genes. We fold-change on the horizontal axis against used RNAhybrid to determine the the statistical significance plotted as -log of microRNA:mRNA pairing with the lowest the p value on the vertical axis (Figure 3A). free energy for each repressed target. Next, Targets were considered significant if the - the predicted miR-106b binding sites were log p value was above 3.35 (e.g., p < 4.2 x analyzed using MEME Suite to identify the 10-4) and these points are plotted in red for most common motif. The resulting decreased or blue for increased expression. sequence motif includes 6 nucleotides that Target gene expression changes of align with the predicted miR-106b binding significant genes ranged between -1.16 to - site from bases 5-10 (Figure 2G; upper). 2.22 fold reduced expression and +1.15 to This sequence does not include the triplet U +1.47 fold increased expression. Selected at positions 2-4, possibly because A:T pairs targets are indicated (for a complete list of have less favorable free energy than G:C the significant targets, see Table 2). Known pairs and the query set was limited by the target genes RB1 and IL-8 were significantly lowest free energy binding site from repressed [10, 29]. Among the novel targets RNAhybrid. identified in our genome-wide analysis were members of the Kruppel-like factor family, Comparison to CLASH data KLF2 and KLF6. These targets were Supplemental data from [19] included 143 significantly repressed by miR-106b with a p target mRNAs bound by miR-106b, value of 1.72*10-7 and 2.85*10-4, including the sequence at the region of respectively. interaction. To validate the miR-106b We compared our experimentally binding characteristics we have described, determined set of 112 mRNAs repressed by we searched for sequence motifs in the miR-106b to the genes predicted by CLASH data set using MEME Suite. Within TargetScan [30] or Micro-T [31], and the 143 sequences, 138 contained a compared the two prediction programs to sequence highly related to each other as well. Figure 3B shows a Venn 5’CTGTCAGCACTTTC. This is the diagram of the number of shared targets in complement of the 5’ end of miR-106b from these data sets. The majority of genes bases 2-14 (with ‘C’ opposite the first contained in our data set were predicted by position of miR-106b slightly favored over one or both programs, though 35 the expected ‘A’). The most conserved experimentally identified targets were not region of this meme is 5’CAGCA, predicted by either program (Figure 3B). complementary to miR-106b bases 6-10 We have also compared our target list with (Figure 2G; lower). Comparing the two the genes confirmed by CLASH and find motifs, the sequence from bases 6-10 is that only ERO1L, FAM91A1, and YES1 most important with some contribution from were identified as miR-106b targets both in flanking bases and tolerance for substitution our data and that of Tollervey and opposite position 5 (allowing a G:U wobble). colleagues [19]. The two data sets are mostly in agreement and indicate that binding of miR-106b is Target validation

We sought to verify that miR-106b levels Based on the observed decrease in KLF2 altered expression of target genes identified and KLF6 in the RNA-Seq data set, we by RNA-Seq. We transfected Mz-ChA-1 tested the effect of miR-106b on additional cells with negative control LNA, miR-106b, KLF family members. KLFs represent a or LNA-106b and isolated total RNA. large family of transcription factors of which Transcript levels of selected targets were many act as tumor suppressors and are determined by quantitative RT-PCR and often down-regulated in cancer [32, 33]. We normalized to 18S rRNA. Targets confirmed confirmed that KLF2 and KLF6 mRNAs by RT-PCR include EREG, RRM2, ITGA2, were regulated by miR-106b and found RB1, GLO1, M6PR, and PSD3. We also other KLF family members KLF4, KLF10, evaluated non-target negative controls KLF11 and KLF13 to have increased ITGA3 and HRAS and found no change by expression when cells were transfected with LNA-106b compared to negative control LNA-106b (Figure 5A). Additionally, we LNA. (Figure 4A). NCEH1 was a target examined the effects of miR-106b on identified by RNA-Seq which had a trend protein expression of KLFs by immunoblot towards increased mRNA expression upon in Mz-ChA-1 cells. We observed decreased miR-106b antagonism, but was not expression of KLF2, KLF4, KLF6, and statistically significant (p = 0.07). We did not KLF10 after 24 hours of transfection with observe a change in mRNA level in RNA- miR-106b compared to negative control Seq target FOS by RT-PCR. Not all miR- LNA and a slight increase in expression with 106b targets will be identified by the current LNA-106b (Figure 5B). method, specifically those that do not have Proliferation a decrease in mRNA levels. We have not exhaustively tried to determine the identity To investigate the role of miR-106b in of such targets, but did recognize the TRAIL proliferation of cholangiocarcinoma cells, we death receptor DR5 as a predicted miR- assessed the change in cell number over 106b target by TargetScan. Because miR- time using the MTT assay. Mz-ChA-1, 106b is clustered with miR-25 and miR-25 KMCH, and BDEneu cholangiocarcinoma regulates cell death by targeting (DR4) [4], cells were transfected with miR-106b, LNA- we tested whether miR-106b decreased the 106b or negative control LNA for 24 hours. functionally-related DR5 protein. Cells were then allowed to grow for up to 72 Transfection with miR-106b decreased DR5 hours. We did not observe any significant protein levels (Figure 4B), potentially acting difference in cell proliferation upon alteration to complement miR-25-mediated DR4 of miR-106b levels (Figure 6). To eliminate repression and promote TRAIL resistance. the possibility of a long-term effect on cell growth we repeated the assay over a one We used biotinylated miR-106b in order to week course in Mz-ChA-1 cells and again validate targets by affinity binding and observed no change in proliferation (data capture. Mz-ChA-1 cells were transfected not shown). with either biotinylated miR-106b or biotinylated C. elegans miR-67 (Cel-miR-67) as a control. Biotin-bound RNA was isolated and RT-PCR was performed for IL-8 and KLF2. We observed significant enrichment of IL-8 and KLF2 transcripts with biotinylated miR-106b pulldown versus control Cel-miR-67 pulldown (Figure 4C). miR-106b targets multiple KLF family members

miR-106b protects against apoptosis transfected with miR-106b, LNA-106b or negative control followed by treatment with KLF2, KLF6 and KLF10 were all regulated either TRAIL or staurosporine to induce cell by miR-106b and are known to promote death. We observed a decrease in apoptotic apoptosis [34-36]. Our data demonstrated nuclei by DAPI staining upon transfection additionally that DR5, a pro-apoptotic death with miR-106b and an increase in apoptotic receptor, was regulated by miR-106b. Thus, nuclei upon transfection with LNA-106b we reasoned that miR-106b may protect (Figure 7). Thus, miR-106b acts in part to against apoptosis in cholangiocarcinoma protect cholangiocarcinoma cells from cells. H69, KMCH, and Mz-ChA-1 cells were apoptosis.

Discussion MicroRNA binding depends on sequence complementarity. The degree of The data presented in this manuscript relate complementarity and length of consecutive to mRNAs regulated by miR-106b in bases can vary, resulting in refinement of cholangiocarcinoma cells. A number of rules of binding [37] and definition of new cancer types over express miR106b, types or classes of microRNA:mRNA making a target gene set of potential value interactions [19]. Classes I-III of interactions to tumors beyond cholangiocarcinoma. The involve complementarity within the seed principle findings reported here show: (i) region. Class IV interactions show miR-106b repressed 112 mRNA targets; (ii) complementarity in a more central region, most target genes contained a 7- or 8-mer described as centered pairing [18]. Finally, seed-binding site; (iii) multiple KLF family Class V interactions exhibited distributed proteins are targeted by miR-106b; and (iv) pairing, where discrete continuous regions miR-106b promoted tumor cell survival in of complementarity were not observed [19]. cholangiocarcinoma cells. Each of these These data are consistent with a role for findings will be discussed below. microRNAs in regulating expression of Our study has revealed 112 mRNAs that mRNAs based on sequence were negatively regulated by miR-106b. complementarity but not strictly limited to Some known miR-106b-regulated genes seed pairing. Perhaps not surprisingly, the (e.g., RB1, IL-8, F3, YES1, FAM91A1, and GC content of microRNA binding motifs, ERO1L) were identified and several novel representing the average or commonly targets were uncovered as well. Our identified binding site over many mRNAs, experiments did not further investigate a was higher than the GC content of functional role for these known targets. Not microRNA seed regions in general [19]. all previously-identified miR-106b-regulated Thus, binding energy may be more genes were significantly altered in our important than binding position, a concept study. Specifically, we did not observe any incorporated into the microRNA target change in the mRNA levels of PTEN, E2F1, prediction algorithm RNAhybrid [25]. or BCL2L11 (Bim). A lack of change in Over 70% of mRNAs that were decreased expression of these mRNAs may reflect contained either a 7-mer or 8-mer miR-106b cell-line specific differences in microRNA binding site. Analysis of all 8-mers or 7- targeting, changes in mRNA levels below mers in miR-106b-regulated sequences the threshold of detection, or post- demonstrated that the sequences at or near transcriptional effects that do not change the seed, including up to nucleotide 10, mRNA levels. Previously unknown genes were over-represented. This over- that were regulated at the mRNA level representation of the 7-mer and 8-mer include KLF family members, which are sequences was highly statistically significant discussed below. when compared to the expected distribution of the same sequences in a thousand

random gene sets of the same size. The FAM91A1 each have a single copy of the 7- seed sequence tolerated a ‘U’ in the place mer-m8 binding site and YES1 has an of ‘C’ (resulting in G:U wobble pairing) additional 7-mer-A1 binding site. The cell opposite positions 5, 7, or both. Two lines used in the two studies are very hydrogen bonds connect the G:U pair while different, Flp-In T-REx 293 embryonic three hydrogen bonds stabilize the G:C pair, kidney-derived cells versus Mz-ChA-1 suggesting there may be a thermodynamic biliary tract cancer-derived cells. The cost to miR-106b binding sites with G:U techniques used are also different. Finally, wobbles. Alternatively, the exocyclic amino data from the CLASH study reflect steady- group of ‘G’ is available for additional state interactions of all detected microRNAs interactions when ‘G’ is paired with ‘U’ [38] and their targets, while the current study allowing for the possibility of compensatory assessed acute changes to mRNA targets stabilizing hydrogen bonds to functional after manipulation of only miR-106b. It is groups within the RISC polypeptides. likely that different cell types will have a different microRNA target landscape, and Based on both the CLASH dataset [19] and that identification of these targets by several our own, we found that the sequence methods will allow a detailed understanding 5’UCAGCACU represents an ‘average’ of mRNA regulation and binding by sequence motif serving as a binding site for microRNAs. miR-106b, with the best evidence for the central 6-mer (underlined, complementary An important finding in the current study to miR-106b based 5-10). While this is the was the coordinated regulation of multiple average binding sequence, the most KLF family members. The seventeen prevalent 7-mer was 5’GCACUUU and the members of the KLF family of transcription most common 8-mer was 5’AGCACUUU. factors are involved in a diverse range of The difference between the average and the biological functions and aberrations can most prevalent sequences is that the lead to disorders such as cardiovascular average (in our data set) was determined and respiratory disease, obesity, using the single-most-stable predicted inflammatory diseases and cancer [39]. binding site, as identified by RNAhybrid. Many KLF family members have been Such a filter will bias against the triplet implicated in some aspect of cancer cell UUU. Still, the agreement between our biology including growth, apoptosis, average binding motif and the motif differentiation, and migration [40]. We have generated from CLASH data where this revealed regulation of six KLF members by potential bias is not relevant suggest that miR-106b in our study. Two members, KLF2 this filter is not unreasonable. Overall, we and KLF6, were significantly repressed find good evidence that binding favors genes in our RNA-Seq experiment. complementarity near the 5’ end of the Additionally, KLF4, KLF10, and KLF11 were microRNA, consistent with the seed model, somewhat near the cutoff for significance (p as well as evidence that the sequence = 0.024, 0.038, and 0.032 respectively), tolerates a slight shift toward the center of while KLF13 mRNA in the RNA-Seq data miR-106b. did not suggest regulation (p = 0.81). These six KLFs were demonstrated to be miR- Comparison of our data set of modulated 106b-responsive genes by qRT-PCR. In genes and that obtained by CLASH showed particular, the result for KLF13 was a striking near-absence of overlap in the surprising as this gene was included as a mRNAs identified. Indeed, of the 143 control under the expectation it would not be mRNAs in the CLASH set and 112 genes responsive to miR-106b. The coordinated down regulated in our experiment, only modulation of these six family members three mRNAs were on both lists: ERO1L, may indicate a functional aspect of miR- YES1, and FAM91A1. None of these 106b biology that was not previously contain an 8-mer binding site. YES1 and

appreciated. Each of the KLFs regulated by Resistance to apoptosis is a characteristic miR-106b in our study has tumor of cholangiocarcinoma cells. Many KLFs are suppressive function in one or several implicated in regulation of apoptosis in cancers. KLF2 has been shown to induce tumor cells but their role in apoptosis and to be a tumor suppressor in cholangiocarcinoma cell death is unknown. prostate and breast cancer cell lines as well We have shown regulation of six KLFs by as in xenografted mice [33]. In pancreatic miR-106b. Overexpression of KLF2 in cancer cells, KLF2 expression is decreased hepatocellular carcinoma cell lines led to and its enforced expression leads to increased cell death [49]. KLF6 has dual inhibition of growth and metastasis [41]. roles in apoptosis as its wildtype form is pro- KLF4 regulates proliferation and apoptotic [50] while the splice variant SV1 differentiation of lung cancer cells and its which is overexpressed in cancer [51] is deletion in a mouse is enough to generate anti-apoptotic. KLF10 promotes cell death in tumors [42]. KLF6 reduced tumorigenic human leukemia cells through upregulation features in osteosarcoma cells [43] and its of pro-apoptotic proteins Bim and Bax [36]. expression is reduced in human and mouse In our study, antagonism of miR-106b with prostate cancer [44]. Loss of KLF6 LNA led to increased apoptosis sensitivity in expression results in increased liver mass, cholangiocarcinoma cells. In part, this effect decreased cyclin-dependent kinase inhibitor could be through derepression of pro- p21, and correlated with low p21 in liver apoptotic KLFs. Targeting of miR-106b to cancer [45]. Mice deficient in KLF10 exhibit increase cholangiocarcinoma cell sensitivity increased skin tumorigenesis when exposed to apoptosis is a potential future strategy. to carcinogens [46]. Knockdown of KLF11 in In summation, we report a landscape of leiomyoma cells leads to increased miR-106b responsive genes in proliferation and its expression is lower in cholangiocarcinoma cells. Several tumor tumor tissue compared to normal [47]. suppressive members of the KLF family of KLF13 is shown to repress anti-apoptotic transcription factors were revealed to be Bcl-2 in acute lymphoblastic leukemia [48]. modulated by miR-106b. And finally, miR- These varied and overlapping anti- 106b is protective against apoptosis in tumorigenic features of KLFs in cancer cholangiocarcinoma cells. make them attractive potential targets for future study in cholangiocarcinoma.

Funding This work was supported by the Center for Drug Delivery and Nanomedicine, UNMC, funded by the National Institutes of Health [P20GM103480]. The PI received support from the Fred & Pamela Buffett Cancer Center Support Grant [P30CA036727]. CJW was supported by a UNMC Graduate Student Fellowship. The Advanced Microscopy Core Facility is supported in part by the Cancer Center as well as the Nebraska Center for Cellular Signaling [P30GM106397]. The High-Throughput DNA Sequencing and Genotyping Core Facility receives partial support from the NCRR [1S10RR027754-01, 5P20RR016469, RR018788-08] and the NIGMS [8P20GM103427, P30GM110768]. The authors are solely responsible for the contents of this publication; it does not necessarily represent the official views of the NIH or NIGMS.

References

1. Meng, F., R. Henson, M. Lang, H. Wehbe, S. Maheshwari, J.T. Mendell, J. Jiang, T.D. Schmittgen, and T. Patel, Involvement of human micro‐RNA in growth and response to chemotherapy in human cholangiocarcinoma cell lines. Gastroenterology, 2006. 130(7): p. 2113‐29. 2. Kawahigashi, Y., T. Mishima, Y. Mizuguchi, Y. Arima, S. Yokomuro, T. Kanda, O. Ishibashi, H. Yoshida, T. Tajiri, and T. Takizawa, MicroRNA profiling of human intrahepatic cholangiocarcinoma cell lines reveals biliary epithelial cell‐specific microRNAs. J Nippon Med Sch, 2009. 76(4): p. 188‐97. 3. Selaru, F.M., A.V. Olaru, T. Kan, S. David, Y. Cheng, Y. Mori, J. Yang, B. Paun, Z. Jin, R. Agarwal, et al., MicroRNA‐21 is overexpressed in human cholangiocarcinoma and regulates programmed cell death 4 and tissue inhibitor of metalloproteinase 3. Hepatology, 2009. 49(5): p. 1595‐601. 4. Razumilava, N., S.F. Bronk, R.L. Smoot, C.D. Fingas, N.W. Werneburg, L.R. Roberts, and J.L. Mott, miR‐25 targets TNF‐related apoptosis inducing ligand (TRAIL) death receptor‐4 and promotes apoptosis resistance in cholangiocarcinoma. Hepatology, 2012. 55(2): p. 465‐75. 5. Gupta, S., D.E. Read, A. Deepti, K. Cawley, A. Gupta, D. Oommen, T. Verfaillie, S. Matus, M.A. Smith, J.L. Mott, et al., Perk‐dependent repression of miR‐106b‐25 cluster is required for ER stress‐induced apoptosis. Cell Death Dis, 2012. 3: p. e333. 6. Kim, K., G. Chadalapaka, S.S. Pathi, U.H. Jin, J.S. Lee, Y.Y. Park, S.G. Cho, S. Chintharlapalli, and S. Safe, Induction of the transcriptional repressor ZBTB4 in prostate cancer cells by drug‐induced targeting of microRNA‐17‐92/106b‐25 clusters. Mol Cancer Ther, 2012. 11(9): p. 1852‐62. 7. Tsujiura, M., D. Ichikawa, S. Komatsu, A. Shiozaki, H. Takeshita, T. Kosuga, H. Konishi, R. Morimura, K. Deguchi, H. Fujiwara, et al., Circulating microRNAs in plasma of patients with gastric cancers. Br J Cancer, 2010. 102(7): p. 1174‐9. 8. Shen, G., H. Jia, Q. Tai, Y. Li, and D. Chen, miR‐106b downregulates adenomatous polyposis coli and promotes cell proliferation in human hepatocellular carcinoma. Carcinogenesis, 2013. 34(1): p. 211‐9. 9. Petrocca, F., R. Visone,R. M. Onelli, M.H. Shah, M.S. Nicoloso, I. de Martino, D. Iliopoulos, E. Pilozzi, C.G. Liu, M. Negrini, et al., E2F1‐regulated microRNAs impair TGFbeta‐dependent cell‐ cycle arrest and apoptosis in gastric cancer. Cancer Cell, 2008. 13(3): p. 272‐86. 10. Cai, K., Y. Wang, and X. Bao, MiR‐106b promotes cell proliferation via targeting RB in laryngeal carcinoma. J Exp Clin Cancer Res, 2011. 30: p. 73. 11. Yang, T.S., X.H. Yang, X. Chen, X.D. Wang, J. Hua, D.L. Zhou, B. Zhou, and Z.S. Song, MicroRNA‐ 106b in cancer‐associated fibroblasts from gastric cancer promotes cell migration and invasion by targeting PTEN. FEBS Lett, 2014. 588(13): p. 2162‐9. 12. Kan, T., F. Sato, T. Ito, N. Matsumura, S. David, Y. Cheng, R. Agarwal, B.C. Paun, Z. Jin, A.V. Olaru, et al., The miR‐106b‐25 polycistron, activated by genomic amplification, functions as an oncogene by suppressing p21 and Bim. Gastroenterology, 2009. 136(5): p. 1689‐700. 13. Vinther, J., M.M. Hedegaard, P.P. Gardner, J.S. Andersen, and P. Arctander, Identification of miRNA targets with stable isotope labeling by amino acids in cell culture. Nucleic Acids Res, 2006. 34(16): p. e107. 14. Baek, D., J. Villen, C. Shin, F.D. Camargo, S.P. Gygi, and D.P. Bartel, The impact of microRNAs on protein output. Nature, 2008. 455(7209): p. 64‐71. 15. Selbach, M., B. Schwanhausser, N. Thierfelder, Z. Fang, R. Khanin, and N. Rajewsky, Widespread changes in protein synthesis induced by microRNAs. Nature, 2008. 455(7209): p. 58‐63. 16. Guo, H., N.T. Ingolia, J.S. Weissman, and D.P. Bartel, Mammalian microRNAs predominantly act to decrease target mRNA levels. Nature, 2010. 466(7308): p. 835‐40.

17. Lim, L.P., N.C. Lau, P. Garrett‐Engele, A. Grimson, J.M. Schelter, J. Castle, D.P. Bartel, P.S. Linsley, and J.M. Johnson, Microarray analysis shows that some microRNAs downregulate large numbers of target mRNAs. Nature, 2005. 433(7027): p. 769‐73. 18. Shin, C., J.W. Nam, K.K. Farh, H.R. Chiang, A. Shkumatava, and D.P. Bartel, Expanding the microRNA targeting code: functional sites with centered pairing. Mol Cell, 2010. 38(6): p. 789‐ 802. 19. Helwak, A., G. Kudla, T. Dudnakova, and D. Tollervey, Mapping the human miRNA interactome by CLASH reveals frequent noncanonical binding. Cell, 2013. 153(3): p. 654‐65. 20. Knuth, A., H. Gabbert, W. Dippold, O. Klein, W. Sachsse, D. Bitter‐Suermann, W. Prellwitz, and K.H. Meyer zum Buschenfelde, Biliary adenocarcinoma. Characterisation of three new human tumor cell lines. J Hepatol, 1985. 1(6): p. 579‐96. 21. Murakami, T., H. Yano, M. Maruiwa, S. Sugihara, and M. Kojiro, Establishment and characterization of a human combined hepatocholangiocarcinoma cell line and its heterologous transplantation in nude mice. Hepatology, 1987. 7(3): p. 551‐6. 22. Lai, G.H., Z. Zhang, X.N. Shen, D.J. Ward, J.L. Dewitt, S.E. Holt, R.A. Rozich, D.C. Hixson, and A.E. Sirica, erbB‐2/neu transformed rat cholangiocytes recapitulate key cellular and molecular features of human bile duct cancer. Gastroenterology, 2005. 129(6): p. 2047‐57. 23. Trapnell, C., A. Roberts, L. Goff, G. Pertea, D. Kim, D.R. Kelley, H. Pimentel, S.L. Salzberg, J.L. Rinn, and L. Pachter, Differential gene and transcript expression analysis of RNA‐seq experiments with TopHat and Cufflinks. Nat Protoc, 2012. 7(3): p. 562‐78. 24. Bartonicek, N. and A.J. Enright, SylArray: a web server for automated detection of miRNA effects from expression data. Bioinformatics, 2010. 26(22): p. 2900‐1. 25. Rehmsmeier, M., P. Steffen, M. Hochsmann, and R. Giegerich, Fast and effective prediction of microRNA/target duplexes. RNA, 2004. 10(10): p. 1507‐17. 26. Bailey, T.L., M. Boden, F.A. Buske, M. Frith, C.E. Grant, L. Clementi, J. Ren, W.W. Li, and W.S. Noble, MEME SUITE: tools for motif discovery and searching. Nucleic Acids Res, 2009. 37(Web Server issue): p. W202‐8. 27. Lal, A., M.P. Thomas, G. Altschuler, F. Navarro, E. O'Day, X.L. Li, C. Concepcion, Y.C. Han, J. Thiery, D.K. Rajani, et al., Capture of microRNA‐bound mRNAs identifies the tumor suppressor miR‐34a as a regulator of growth factor signaling. PLoS Genet, 2011. 7(11): p. e1002363. 28. Wehrkamp CJ, G.A., Natarajan SK, Phillippi MA, Mott JL., XIAP antagonist embelin inhibited proliferation of cholangiocarcinoma cells. PLoS One, 2014. 29. Chuang, T.D., X. Luo, H. Panda, and N. Chegini, miR‐93/106b and their host gene, MCM7, are differentially expressed in leiomyomas and functionally target F3 and IL‐8. Mol Endocrinol, 2012. 268(6): p. 102 ‐42. 30. Lewis, B.P., I.H. Shih, M.W. Jones‐Rhoades, D.P. Bartel, and C.B. Burge, Prediction of mammalian microRNA targets. Cell, 2003. 115(7): p. 787‐98. 31. Reczko, M., M. Maragkakis, P. Alexiou, I. Grosse, and A.G. Hatzigeorgiou, Functional microRNA targets in protein coding sequences. Bioinformatics, 2012. ):28(6 p. 771‐6. 32. Kremer‐Tal, S., H.L. Reeves, G. Narla, S.N. Thung, M. Schwartz, A. Difeo, A. Katz, J. Bruix, P. Bioulac‐Sage, J.A. Martignetti, et al., Frequent inactivation of the tumor suppressor Kruppel‐like factor 6 (KLF6) in hepatocellular carcinoma. Hepatology, 2004. 40(5): p. 1047‐52. 33. Taniguchi, H., F.V. Jacinto, A. Villanueva, A.F. Fernandez, H. Yamamoto, F.J. Carmona, S. Puertas, V.E. Marquez, Y. Shinomura, K. Imai, et al., Silencing of Kruppel‐like factor 2 by the histone methyltransferase EZH2 in human cancer. Oncogene, 2012. 31(15): p. 1988‐94. 34. Wang, F., Y. Zhu, Y. Huang, S. McAvoy, W.B. Johnson, T.H. Cheung, T.K. Chung, K.W. Lo, S.F. Yim, M.M. Yu, et al., Transcriptional repression of WEE1 by Kruppel‐like factor 2 is involved in DNA damage‐induced apoptosis. Oncogene, 2005. 24(24): p. 3875‐85.

35. D'Astolfo, D.S., R.C. Gehrau, J.L. Bocco, and N.P. Koritschoner, Silencing of the transcription factor KLF6 by siRNA leads to cell cycle arrest and sensitizes cells to apoptosis induced by DNA damage. Cell Death Differ, 2008. 15(3): p. 613‐6. 36. Jin, W., G. Di, J. Li, Y. Chen, W. Li, J. Wu, T. Cheng, M. Yao, and Z. Shao, TIEG1 induces apoptosis through mitochondrial apoptotic pathway and promotes apoptosis induced by homoharringtonine and velcade. FEBS Lett, 2007. 581(20): p. 3826‐32. 37. Grimson, A., K.K. Farh, W.K. Johnston, P. Garrett‐Engele, L.P. Lim, and D.P. Bartel, MicroRNA targeting specificity in mammals: determinants beyond seed pairing. Mol Cell, 2007. 27(1): p. 91‐105. 38. Knitt, D.S., G.J. Narlikar, and D. Herschlag, Dissection of the role of the conserved G.U pair in group I RNA self‐splicing. Biochemistry, 1994. 33(46): p. 13864‐79. 39. McConnell, B.B. and V.W. Yang, Mammalian Kruppel‐like factors in health and diseases. Physiol Rev, 2010. 90(4): p. 1337‐81. 40. Tetreault, M.P., Y. Yang, and J.P. Katz, Kruppel‐like factors in cancer. Nat Rev Cancer, 2013. 13(10): p. 701‐13. 41. Zhang, D., Y. Dai, Y. Cai, T. Suo, H. Liu, Y. Wang, Z. Cheng, and H. Liu, KLF2 is downregulated in pancreatic ductal adenocarcinoma and inhibits the growth and migration of cancer cells. Tumour Biol, 2015. 42. Yu, T., X. Chen, W. Zhang, J. Liu, R. Avdiushko, D.L. Napier, A.X. Liu, J.M. Neltner, C. Wang, D. Cohen, et al., KLF4 regulates adult lung tumor‐initiating cells and represses K‐Ras‐mediated lung cancer. Cell Death Differ, 2016. 23(2): p. 207‐15. 43. Jianwei, Z., B. Enzhong, L. Fan, L. Jian, and A. Ning, Effects of Kruppel‐like factor 6 on osteosarcoma cell biological behavior. Tumour Biol, 2013. 34(2): p. 1097‐105. 44. Chiam, K., N.K. Ryan, C. Ricciardelli, T.K. Day, G. Buchanan, A.M. Ochnik, K. Murti, L.A. Selth, B. Australian Prostate Cancer, L.M. Butler, et al., Characterization of the prostate cancer susceptibility gene KLF6 in human and mouse prostate cancers. Prostate, 2013. 73(2): p. 182‐93. 45. Narla, G., S. Kremer‐Tal, N. Matsumoto, X. Zhao, S. Yao, K. Kelley, M. Tarocchi, and S.L. Friedman, In vivo regulation of p21 by the Kruppel‐like factor 6 tumor‐suppressor gene in mouse liver and human hepatocellular carcinoma. Oncogene, 2007. 26(30): p. 4428‐34. 46. Song, K.D., D.J. Kim, J.E. Lee, C.H. Yun, and W.K. Lee, KLF10, transforming growth factor‐beta‐ inducible early gene 1, acts as a tumor suppressor. Biochem Biophys Res Commun, 2012. 419(2): p. 388‐94. 47. Yin, P., Z. Lin, S. Reierstad, J. Wu, H. Ishikawa, E.E. Marsh, J. Innes, Y. Cheng, K. Pearson, J.S.t. Coon, et al., Transcription factor KLF11 integrates progesterone receptor signaling and proliferation in uterine leiomyoma cells. Cancer Res, 2010. 70(4): p. 1722‐30. 48. Jing, D., V.A. Bhadri, D. Beck, J.A. Thoms, N.A. Yakob, J.W. Wong, K. Knezevic, J.E. Pimanda, and R.B. Lock, Opposing regulation of BIM and BCL2 controls glucocorticoid‐induced apoptosis of pediatric acute lymphoblastic leukemia cells. Blood, 2015. 125(2): p. 273‐83. 49. Huang, M.D., W.M. Chen, F.Z. Qi, M. Sun, T.P. Xu, P. Ma, and Y.Q. Shu, Long non‐coding RNA TUG1 is up‐regulated in hepatocellular carcinoma and promotes cell growth and apoptosis by epigenetically silencing of KLF2. Mol Cancer, 2015. 14: p. 165. 50. Huang, X., X. Li, and B. Guo, KLF6 induces apoptosis in prostate cancer cells through up‐ regulation of ATF3. J Biol Chem, 2008. 283(44): p. 29795‐801. 51. Chen, H., L. Chen, L. Sun, H. Zhen, X. Li,d an Q. Zhang, A small interfering RNA targeting the KLF6 splice variant, KLF6‐SV1, as gene therapy for gastric cancer. Gastric Cancer, 2011. 14(4): p. 339‐ 52.

Table 1: Primers Used for qRT-PCR

Gene Name Forward Primer (5’ -> 3’) Reverse Primer (5’ -> 3’) Retinoblastoma 1 GTCGTTCACTTTTACTGAGC TCCAATTTGCTGAAGAGTGC Integrin, alpha 2 GAGTGGCTTTCCTGAGAACC CTGGTGAGGATCAAGCCGAG Mannose-6-phosphate receptor GATTCTGAGCTTTGGCTACT GTCTGCCAGGATTCTCTCAC Pleckstrin and Sec7 domain CTGGCGATGGAAGATGGAAG CATATTTGGCCTTGGCAACAC containing 3 Glyoxalase 1 GATACTGCAGCGCAGCCATG CCAGTGACTTCTTAGGATCC Epiregulin GTCCTCAGTACAACTGTGAT GACACTTGAGCCACACGTGG Ribonucleotide reductase M2 TCTGCTTCGCTGCGCCTCCA TGGAAGATCCTCCTCGCGGT Harvey rat sarcoma viral CCATCCAGCTGATCCAGAAC TGTCCAACAGGCACGTCTC oncogene homolog (HRAS) Integrin, alpha 3 TGCGTCGTCTCCGCCTTCAA CATCCGCTCACAGTCATCCT Neutral cholesterol ester GAAGCTGATGCTGCTGGACG AACACTCTGACTTCCACACC hydrolase 1 FBJ murine osteosarcoma viral GCCTAACCGCCACGATGATG GGACTGGTCGAGATGGCAGT oncogene homolog (FOS) Interleukin 8 CAAGAGCCAGGAAGAAACCA ATTTGGGGTGGAAAGGTTTG KLF2 ACTCACACCTGCAGCTACGC GCACAGATGGCACTGGAAT KLF4 CAGAGGAGCCCAAGCCAAAG CCAGTCACAGTGGTAAGGTT KLF6 CTGCCGTCTCTGGAGGAGT TCCACAGATCTTCCTGGCTGTC KLF10 ACCAAACGAGTCTGGACAGT TCAGATACTGGTGTAACAGG KLF11 ACTGTGCATATGGATGCAGC TACGGCAGAGGACTGGAGAA KLF13 CTCAAGGCGCACCTGAGAAC GTCAGGTGGTCGCTGCGCAT 18S CGTTCTTAGTTGGTGGAGCG CGCTGAGCCAGTCAGTGTAG bioRxiv preprint doi: https://doi.org/10.1101/088229; this version posted November 17, 2016. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

Fold log2(fold Table 2: Gene Name, Down-regulated Change change) tubulin, alpha 1a (TUBA1A), transcript variants 2, 3, and 1 0.45 -1.16 pleckstrin and Sec7 domain containing 3 (PSD3), transcript variants 1 and 2 0.64 -0.65 Kruppel-like factor 2 (lung) (KLF2) 0.65 -0.63 centromere protein Q (CENPQ) 0.66 -0.60 interleukin 8 (IL8) 0.67 -0.58 cyclin E2 (CCNE2) 0.69 -0.53 abhydrolase domain containing 13 (ABHD13) 0.69 -0.53 transmembrane protein 64 (TMEM64), transcript variants 1 and 2 0.70 -0.52 epiregulin (EREG) 0.71 -0.50 osteopetrosis associated transmembrane protein 1 (OSTM1) 0.71 -0.50 BTG family, member 3 (BTG3), transcript variants 1 and 2 0.71 -0.50 FBJ murine osteosarcoma viral oncogene homolog (FOS) 0.71 -0.50 thioredoxin-related transmembrane protein 3 (TMX3) 0.72 -0.48 zinc finger protein 367 (ZNF367) 0.72 -0.48 LysM, putative peptidoglycan-binding, domain containing 3 (LYSMD3) 0.73 -0.46 cathepsin S (CTSS), transcript variants 2 and 1 0.73 -0.45 chromosome 2 open reading frame 69 (C2orf69) 0.73 -0.44 mitogen-activated protein kinase kinase kinase 2 (MAP3K2) 0.74 -0.44 FtsJ methyltransferase domain containing 1 (FTSJD1), transcript variants 2 and 1 0.74 -0.43 interferon-induced protein with tetratricopeptide repeats 5 (IFIT5) 0.74 -0.43 OTU domain containing 1 (OTUD1) 0.75 -0.42 ArfGAP with coiled-coil, ankyrin repeat and PH domains 2 (ACAP2) 0.76 -0.40 coagulation factor III (thromboplastin, tissue factor) (F3), transcript variants 2 and 1 0.76 -0.39 discoidin, CUB and LCCL domain containing 2 (DCBLD2) 0.76 -0.39 neutral cholesterol ester hydrolase 1 (NCEH1), transcript variants 1, 3, 4, and 2 0.77 -0.39 zinc finger and BTB domain containing 6 (ZBTB6) 0.77 -0.38 early endosome antigen 1 (EEA1) 0.77 -0.38 breast cancer metastasis-suppressor 1-like (BRMS1L) 0.77 -0.37 centrosomal protein 152kDa (CEP152), transcript variants 1 and 2 0.78 -0.37 nuclear protein, ataxia-telangiectasia locus (NPAT) 0.78 -0.36 RAB27B, member RAS oncogene family (RAB27B) 0.78 -0.36 tankyrase, TRF1-interacting ankyrin-related ADP-ribose polymerase 2 (TNKS2) 0.78 -0.36 family with sequence similarity 3, member C (FAM3C), transcript variants 2 and 1 0.78 -0.36 EF-hand calcium binding domain 14 (EFCAB14) 0.78 -0.36 trans-golgi network vesicle protein 23 homolog B (S. cerevisiae) (TVP23B) 0.78 -0.36 RMI1, RecQ mediated genome instability 1, homolog (S. cerevisiae) (RMI1) 0.78 -0.36 receptor accessory protein 3 (REEP3) 0.78 -0.35 transmembrane protein 245 (TMEM245) 0.78 -0.35 TBC1 domain family, member 9 (with GRAM domain) (TBC1D9) 0.79 -0.35 Ewing tumor-associated antigen 1 (ETAA1) 0.79 -0.34 twinfilin, actin-binding protein, homolog 1 (Drosophila) (TWF1), transcript variants 1 and 2 0.79 -0.34 transmembrane protein 167A (TMEM167A) 0.79 -0.33 TruB pseudouridine (psi) synthase homolog 1 (E. coli) (TRUB1) 0.80 -0.33 glyoxalase I (GLO1) 0.80 -0.32 ribonucleotide reductase M2 (RRM2), transcript variant 1 0.80 -0.32 integrin, alpha 2 (CD49B, alpha 2 subunit of VLA-2 receptor) (ITGA2), transcript variant 1 0.80 -0.32 thioredoxin domain containing 9 (TXNDC9) 0.80 -0.31 RAB22A, member RAS oncogene family (RAB22A) 0.80 -0.31 UDP-GlcNAc:betaGal beta-1,3-N-acetylglucosaminyltransferase 5 (B3GNT5) 0.80 -0.31 retinoblastoma 1 (RB1) 0.80 -0.31 ankyrin repeat domain 22 (ANKRD22) 0.81 -0.31 fatty acyl CoA reductase 1 (FAR1) 0.81 -0.31 LIM domain kinase 1 (LIMK1), transcript variants 2 and 1 0.81 -0.31 ubiquitin specific peptidase 1 (USP1), transcript variants 2, 3, and 1 0.81 -0.31 polo-like kinase 2 (PLK2), transcript variants 2 and 1 0.81 -0.31 deoxycytidine kinase (DCK) 0.81 -0.30 receptor accessory protein 5 (REEP5) 0.81 -0.30 CD46 molecule, complement regulatory protein (CD46), transcript variants a,d,n,c,e,f,b,l 0.81 -0.30 aryl hydrocarbon receptor nuclear translocator-like 2 (ARNTL2), transcript variants 2,3,4,5,1 0.81 -0.30 WD repeat domain 36 (WDR36) 0.81 -0.30 paraoxonase 2 (PON2), transcript variants 1 and 2 0.82 -0.29 soc-2 suppressor of clear homolog (C. elegans) (SHOC2), transcript variants 2 and 1 0.82 -0.29 mannose-6-phosphate receptor (cation dependent) (M6PR), transcript variants 2 and 1 0.82 -0.29 protein phosphatase 3, regulatory subunit B, alpha (PPP3R1) 0.82 -0.29 protein phosphatase 1, regulatory subunit 15B (PPP1R15B) 0.82 -0.29 TMED7-TICAM2 readthrough (TMED7-TICAM2), transcript variants 1 and 2 0.82 -0.28 fatty acid binding protein 5 (psoriasis-associated) (FABP5) 0.82 -0.28 transmembrane emp24 protein transport domain containing 5 (TMED5), transcript variants 2 and 1 0.82 -0.28 Kruppel-like factor 6 (KLF6), transcript variants B, C, and A 0.82 -0.28 ciliary neurotrophic factor (CNTF); ZFP91 zinc finger protein (ZFP91), transcript variants 2 and 1 0.82 -0.28 COP9 signalosome subunit 2 (COPS2), transcript variants 2 and 1 0.83 -0.28 dynein, cytoplasmic 1, light intermediate chain 2 (DYNC1LI2) 0.83 -0.28 karyopherin alpha 3 (importin alpha 4) (KPNA3) 0.83 -0.27 structural maintenance of chromosomes flexible hinge domain containing 1 (SMCHD1) 0.83 -0.27 glia maturation factor, beta (GMFB) 0.83 -0.27 dynein, light chain, Tctex-type 3 (DYNLT3) 0.83 -0.27 MLF1 interacting protein (MLF1IP) 0.83 -0.27 dickkopf 1 homolog (Xenopus laevis) (DKK1) 0.83 -0.27 integrin, alpha 6 (ITGA6), transcript variants 2 and 1 0.83 -0.27 bioRxiv preprint doi: https://doi.org/10.1101/088229; this version posted November 17, 2016. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

thioredoxin-related transmembrane protein 1 (TMX1) 0.83 -0.26 YTH domain family, member 3 (YTHDF3) 0.83 -0.26 UDP-N-acetyl-alpha-D-galactosamine:polypeptide N-acetylgalactosaminyltransferase 3 (GalNAc-T3) (GALNT3) 0.83 -0.26 reticulocalbin 2, EF-hand calcium binding domain (RCN2), transcript variants 2 and 1 0.83 -0.26 RAB12, member RAS oncogene family (RAB12) 0.83 -0.26 v-yes-1 Yamaguchi sarcoma viral oncogene homolog 1 (YES1) 0.83 -0.26 eukaryotic translation initiation factor 2, subunit 1 alpha, 35kDa (EIF2S1) 0.84 -0.26 transmembrane protein 123 (TMEM123) 0.84 -0.26 epithelial cell transforming sequence 2 oncogene (ECT2), transcript variants 1, 2, and 3 0.84 -0.25 ATPase family, AAA domain containing 2 (ATAD2) 0.84 -0.25 RNA binding motif (RNP1, RRM) protein 3 (RBM3) 0.84 -0.25 transmembrane protein 30A (TMEM30A), transcript variants 2 and 1 0.84 -0.25 family with sequence similarity 91, member A1 (FAM91A1) 0.84 -0.25 ATP-binding cassette, sub-family E (OABP), member 1 (ABCE1), transcript variants 2 and 1 0.84 -0.25 integrin, alpha V (ITGAV), transcript variants 2, 3 and 1 0.84 -0.25 RAD18 homolog (S. cerevisiae) (RAD18) 0.84 -0.25 cytidine monophosphate (UMP-CMP) kinase 1, cytosolic (CMPK1), transcript variants 2 and 1 0.84 -0.25 signal peptidase complex subunit 3 homolog (S. cerevisiae) (SPCS3) 0.84 -0.24 ERO1-like (S. cerevisiae) (ERO1L) 0.84 -0.24 ras homolog family member B (RHOB) 0.85 -0.24 alkylglycerone phosphate synthase (AGPS) 0.85 -0.24 nuclear fragile X mental retardation protein interacting protein 2 (NUFIP2) 0.85 -0.24 ubiquitin protein ligase E3C (UBE3C) 0.85 -0.24 prostaglandin-endoperoxide synthase 2 (prostaglandin G/H synthase and cyclooxygenase) (PTGS2) 0.85 -0.23 potassium channel tetramerisation domain containing 9 (KCTD9) 0.85 -0.23 poly(A) polymerase alpha (PAPOLA), transcript variants 2, 3, and 1 0.85 -0.23 solute carrier family 25 (mitochondrial carrier; phosphate carrier), member 24 (SLC25A24), transcript variants 1 and 2 0.85 -0.23 eukaryotic translation initiation factor 1A, X-linked (EIF1AX) 0.85 -0.23 nicotinamide phosphoribosyltransferase (NAMPT) 0.85 -0.23 lysophospholipase I (LYPLA1) 0.85 -0.23 DEK oncogene (DEK), transcript variants 2 and 1 0.85 -0.23 CD55 molecule, decay accelerating factor for complement (Cromer blood group) (CD55), transcript variants 1 and 2 0.86 -0.22 solute carrier family 38, member 2 (SLC38A2) 0.86 -0.22 Fold log2(fold Gene Name, Up-regulated Change change) lipocalin 2 (LCN2) 1.16 0.21 FK506 binding protein 8, 38kDa (FKBP8) 1.17 0.23 biliverdin reductase B (flavin reductase (NADPH)) (BLVRB) 1.17 0.23 ribosomal protein S28 (RPS28) 1.18 0.24 pancreatic progenitor cell differentiation and proliferation factor homolog (zebrafish) (PPDPF) 1.19 0.25 phosphoglycerate dehydrogenase (PHGDH) 1.19 0.25 mevalonate (diphospho) decarboxylase (MVD) 1.20 0.26 mucin 5B, oligomeric mucus/gel-forming (MUC5B) 1.21 0.27 ribosomal protein L28 (RPL28), transcript variants 2, 1, 3, 4, and 5 1.21 0.28 (Unassigned; overlap with mucin 5AC, oligomeric mucus/gel-forming (MUC5AC)) 1.22 0.29 N-myc downstream regulated 1 (NDRG1), transcript variants 1,3, 4, and 2 1.24 0.31 RNA, 5.8S ribosomal 5 (RNA5-8S5), ribosomal RNA 1.25 0.32 phosphoenolpyruvate carboxykinase 2 (mitochondrial) (PCK2), transcript variants 2 and 1 1.26 0.33 DNA-damage-inducible transcript 4 (DDIT4) 1.31 0.39 mucin 2, oligomeric mucus/gel-forming (MUC2) 1.36 0.44 ubiquitin domain containing 1 (UBTD1) 1.43 0.52 LY6/PLAUR domain containing 2 (LYPD2) 1.48 0.56 bioRxiv preprint doi: https://doi.org/10.1101/088229; this version posted November 17, 2016. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

Figure legends

Figure 1. miR-106b targets in cholangiocarcinoma cells predominantly contain a seed-binding site. (A) miR-106b RNA levels were measured by qRT-PCR after transfection of Mz-ChA-1 cells with control LNA (Control), antagonist to miR-106b (LNA-106b), or miR-106b. Expression was normalized to Z30 and plotted as relative level. The dashed horizontal line indicates the mean of miR-106b in Control cells. Data are mean ± SEM for three samples each. *** P < 0.001 using ANOVA with post hoc correction. (B) Following RNA-Seq of the LNA-106b and miR-106b samples from panel A, significantly altered transcripts were categorized as increased (13.2%) or decreased (86.8%) by miR-106b. (C) A majority of decreased transcripts contained one or more miR-106b seed-binding sites (7-mer or 8-mer) while only one increased transcript contained a 7-mer binding site. None of the increased transcripts contained an 8-mer binding site (#). (D) All transcripts identified by RNA-Seq were sorted by statistical significance of expression difference between miR-106b and LNA-106b. This sorted list (plotted on the horizontal axis) was then analyzed by SylArray to identify microRNA binding sites that are over-represented, shown by colored traces. Over-representation of seed-binding sites is indicated on the vertical axis above zero, plotted on a log scale. Under-represented sequences are plotted below zero.

Figure 2. Preferential miR-106b target binding via the 5’ end of the microRNA. (A) Histogram of the distibution of 8-mers found in miR-106b-regulated sequences. The horizontal axis depicts the frequency each 8-mer was observed and the vertical axis represents the number of 8-mers at each count. The bins containing the miR-106b 1-8 and 2-9 perfect binding sequences are indicated. (B) Data are as in panel A except that 7- mers were analyzed. Bins containing the miR-106b 1-7, 2-8, and 3-9 perfect binding sequences are indicated.

(C) The count of each possible 8-mer binding site in the miR-106b-regulated sequences is indicated. Sequences are ordered as they occur along miR-106b. The horizontal dashed line represents the count corresponding to the 90th percentile of all 8-mers (i. e., only the top 10 percent occur more frequently). (D) The same plot as in panel C except using 8-mers derived from let-7a. (E) Counts in the miR-106b-regulated sequences of each possible 7-mer along miR-106b are indicated. The horizontal dashed line represents the count corresponding to the 90th percentile of all 7-mers. (F) The same plot as in panel E except using 7-mers derived from let-7a. (G) Single-base substitutions within the 8-base miR-106b binding site were queried for their frequency, compared to the perfect 8-mer (observed 63 times) and plotted as the nucleotide frequency (count) at each position when the other 7 positions were a perfect miR-106b match. For example, the single substitution of a ‘U’ opposite position 5 (forming a G:U wobble) was observed 40 times while the favored ‘C’ (forming a G:C pair) was found 63 times. (H) k-mer analysis of the count of miR-106b binding sites in the miR- 106b-regulated gene set compared to 1,000 randomly chosen similarly-sized gene sets. (I) The binding motif for miR-106b is shown over the sequence of the microRNA (antiparallel). Taller letters indicate greater representation of that nucleotide in determining the motif. The upper motif was generated using RNA-Seq data from the current study. The analysis was performed separately on data from Helwak et al., 2010, shown in the lower motif.

Figure 3. miR-106b target discovery by RNA-Seq. (A) Volcano plot of gene expression by RNA-Seq. miR- 106b RNA levels were altered by transfection of Mz-ChA-1 cells with LNA-106b or miR-106b and the resulting differential expression of all genes was evaluated. Transcripts are plotted as log2 of expression fold change on the horizontal axis versus –log of the p value on the vertical axis. Gene expression change was considered significant at –log p > 3.35 and genes with values above this cutoff are indicated by colored points. Significantly altered transcripts were either decreased (red) or increased (blue). Labeled genes are those that have been evaluated for modulation by miR-106b through additional experiments. (B) Venn diagram demonstrating overlap of our RNA-Seq miR-106b target repression results with microRNA target prediction databases TargetScan and Micro-T. Our data contained 35 novel targets not predicted by either program.

Figure 4. RNA-Seq target validation. (A) qRT-PCR for nine candidate targets from RNA-Seq. Relative expression is significantly increased for LNA-106b compared to miR-106b in six of the genes. There was a trend towards increased expression for LNA-106b compared to miR-106b for NCEH1 but it was not statistically significant (p = 0.07). FOS showed no significant change in expression by qRT-PCR. Dotted line represents expression level for scrambled control LNA. ITGA3 and HRAS are non-target negative control genes which show no significant expression change. (B) Immunoblot showing transfection with miR-106b decreased the functionally-related DR5 protein levels. (C) Schematic of experimental design for capture of mRNA targets using biotinylated microRNA. Briefly, Mz-ChA-1 cells were transfected for 24 hours with either mature human miR-106b or C. elegans miR-67 which had been biotinylated. Cells were lysed and incubated with streptavidin- bound beads to capture biotinylated microRNA and associated mRNAs. Total RNA was isolated and relative expression of target mRNAs KLF2 and IL-8 was measured for enrichment by qRT-PCR. 18S was used as a control RNA. * p < 0.05; ** p < 0.01; *** p < 0.001; using ANOVA with post hoc correction.

Figure 5. miR-106b regulates multiple KLF family members. (A) qRT-PCR for RNA-Seq candidate miR- 106b targets KLF2 and 6 confirmed regulation at the RNA level. mRNA expression was increased by miR-106b antagonism with LNA-106b compared to miR-106b. Additional KLFs 4, 10, 11, and 13 were evaluated and showed the same pattern of expression increase with LNA-106b antagonism compared to miR-106b treatment. (B) Immunoblots showing regulation of KLFs 2, 4, 6 and 10 by miR-106b at the protein level. 24 hour transfection of Mz-ChA-1 or KMCH cells with control LNA, miR-106b, or LNA-106b led to decrease in KLF protein expression by miR-106b compared to control LNA or LNA-106b. Actin was probed as a loading control. * p < 0.05; ** p < 0.01; *** p < 0.001; using ANOVA with post hoc correction.

Figure 6. miR-106b does not affect proliferation in cholangiocarcinoma cells. We tested the effect of miR- 106b on proliferation in BDEneu, KMCH, and Mz-ChA-1 cells by MTT assay. After 24 hour transfection with control LNA, miR-106b or LNA-106b, cells were allowed to grow for up to 72 hours and cell number was measured by absorbance read at 540 nm. We observed no significant difference in cell growth for any cell line. Signal represents the mean (n = 4) +/- SEM.

Figure 7. miR-106b protects against TRAIL- or staurosporine-induced apoptosis in cholangiocarcinoma cells. H69, KMCH or Mz-ChA-1 cells were transfected with control LNA, miR-106b or LNA-106b for 24 hours followed by treatment with either TRAIL or staurosporine to induce apoptosis. We observed a decrease in apoptotic nuclei by DAPI staining upon transfection with miR-106b and an increase in apoptotic nuclei upon transfection with LNA-106b. DAPI-positive nuclei were counted and expressed as a percent of total nuclei. Data are a mean of three experiments +/- SEM. * p < 0.05; ** p < 0.01; *** p < 0.001; using ANOVA with post hoc correction.